Hi, I would like some help please, if it is possible.
I have to start programming in CUDA, and I have thought in this computer, is ok for starting? or can you recommend me a better one, same price…
I`m interested en this one, because, a friend is selling it, but I don’t know if it will really be useful.
will a have problems with the video drivers in linux?? or is it better to use windows in this case, i have looked for information, but i haven’t found to much for a good start point in CUDA programmig, I mean, a good configuration in hardware, it always talks about good configuration for games.
The apps i have to do, are about live stream and image processing, using GPU off course…
I really want to be sure before spending some money in this…
Please help. I would really appreciate it.
My native language is not English, sorry if I committed some mistakes in my writing.
Looks like a good starter system to me. Linux should be fine. Most Linux distributions are straightforward installs. If you’re using Ubuntu, it’s a bit trickier but you can follow the instructions here.
To get started, you’re better off starting with libraries as opposed to writing your own kernel code directly. For example, take a look at ArrayFire’s image processing.
I have to tell that 2GB of RAM might not be enough, and you need at least 4GB of RAM with a 1GB Video card (to have input and output buffers), and probably will do better with a 8GB (2x4GB) RAM, to enable to asynchronously read/write data while running kernels.
Yes, I like fractals, but it’s more because I agree with Buddha’s philosophy, (nerd and spiritual side mixed xD).
And yes, my first idea was to do something with fractals, but for a start, I think its more useful some image processing, and probably easier, at least there is more info.
Buy an used PC with nVidia GPU, or add it an entry-level Fermi or Kepler GPU; Alternatively buy an used Mac Mini with GeForce 320M; Then use CUDA remotely from your Mac :)
Notice that I use CUDA and OpenCL this way, from my Mac laptop (Radeon HD6750M), using an iMac (Radeon HD5850M dubbed “HD 5750”), and a PC with 2 Graphic card (HD7750 & GTX 260 until I got my new Kepler GT 640).
I would be wary of using the 320M for CUDA these days. At some point in the past few releases, I think the size of the CUDA context (or perhaps the baseline GPU memory usage of the OS) grew to the point that you can’t initialize a CUDA context on a device with only 256 MB of memory. I can no longer run CUDA programs on my MacBook Pro 13" with the GeForce 320M…
Ouch! It’s worth to be mentionned and documented, because many cuda-enabled devices have only 256MB or 128MB, when we think about IGP and older Geforce 8 or GeForce 9 cards!
Well on older GPUs with reduced hardware and compute capabilities I don’t see the need to go with the latest CUDA SDK versions. CUDA 2.3 or CUDA 3.2 should be OK to use - and this also has a bigger compatibility with the installed driver base.